Effects of different wavelet filters on correlation and diagnostic performance of radiomics features / 中南大学学报(医学版)

To investigate the effects of different wavelet filters on correlation and diagnostic performance of radiomics features.
 Methods: A total of 143 colorectal cancer (CRC) patients (64 positive in lymph node metastasis and 79 negative) with contrast-enhanced CT examination were recruited. After labeling the tumor area by experienced radiologists, radiomics wavelets features based on 48 different wavelets were extracted using in-house software coded by Matlab. The correlation coefficients of the features with same names between different wavelets were calculated and got the distribution of high-correlation features between each wavelet. The least absolute shrinkage and selection operator (LASSO) was used to build signatures between lymph node metastasis and wavelet features data set based on different wavelets. The numbers of features in signatures and diagnostic performance were compared using Delong's test.
 Results: With the difference of wavelet order increased, the number of high-correlation features between two wavelets decreased. Some features were prone to high correlation between different wavelets. When building radiomics signature based on single wavelet, signatures built from 'rbio2.2', 'sym7' and 'db7' did well in predicting lymph node metastasis. The signature based on Daubechies wavelet feature set had the highest performance in predicting lymph node metastasis, while the signature from Biorthogonal wavelet features was worst. Improvement was significant in diagnostic performance after excluding the high-correlation features in the whole features set (P=0.004).
 Conclusion: In order to reduce the data redundancy of features, it is recommended to select wavelets with large differences in wavelet orders when calculating radiomics wavelet features. It is necessary to remove high correlation features for improving the diagnostic performance of radiomics signature.

Preoperative Prediction of Lymphovascular Invasion of Colorectal Cancer Based on Radiomics Approach / 中国医学影像学杂志

Purpose Lymph-vascular invasion (LVI) is a risk factor for the prognosis of colorectal cancer, and it is of great value to evaluate the status of lymphatic vessels before treatment. This study aims to predict colorectal cancer LVI preoperatively based on radiomics. Materials and Methods Radiomics features were extracted from preoperative CT images of colorectal cancer retrospectively collected and radiomics labels were constructed. The predictive efficacy of radiomics labels were assessed and internally verified. Joint predictive factors were established by combining clinical factors with independent predictive efficacy and radiomics labels, and their predictive efficacy was evaluated. Results Radiomics labels consisted of 58 radiomics features were correlated with LVI status (P<0.0001)with the former showing good discrimination ability[C-index 0.719,95% CI:0.715-0.723]and classification ability(sensitivity 0.726, specificity 0.628) with internal validation (C-index 0.720). Joint predictive factors containing radiomics labels and carcino-embryonic antigen further enhanced the predictability of radiomics labels (C-index 0.751, sensitivity 0.788, specificity 0.667). Conclusion The radiomics labels built in this study can provide individualized prediction of LVI status of patients with colorectal cancer before surgery. Joint predictive factors in combination with clinical risk factors further improved predictive efficacy.

A CT-based radiomics analysis for clinical staging of non-small cell lung cancer / 中华放射学杂志

Objective To develop and validate a CT-based radiomics predictive model for preoperative predicting the stage of non-small cell lung cancer (NSCLC). Methods In this retrospective study, 657 patients with histologically confirmed was collected from October 2007 to December 2014.The primary dataset consisted of patients with histologically confirmed NSCLC from October 2007 to April 2012, while independent validation was conducted from May 2012 to December 2014.All the patients underwent non-enhanced and contrast-enhanced CT images scan with a standard protocol. The pathological stage (PTNM) of patients with NSCLC were determined by the intraoperative and postoperative pathological findings,and were divided into early stage(Ⅰ,Ⅱstage)and advanced stage(Ⅲ,Ⅳstage).A list of radiomics features were extracted using the software Matlab 2014a and the corresponding radiomics signature was constructed. Multivariable logistic regression analysis was performed with radiomics signature and clinical variables for developing the prediction model. The model performance was assessed with respect to discrimination using the area under the curve (AUC) of receiver operating characteristic(ROC) analysis. Results The discrimination performance of radiomics signature yielded a AUC of 0.715[95% confidence interval (CI):0.709 to 0.721] in the primary dataset and a AUC of 0.724(95% CI:0.717 to 0.731) in the validation dataset. On multivariable logistic regression, radiomics signature, tumor diameter, carcinoembryonic antigen (CEA) level, and cytokeratin 19 fragment (CYFRA21-1) level were showed independently associated with the stage ( Ⅰ,Ⅱ stage vs. Ⅲ, Ⅳ stage) of NSCLC. The prediction model showed good discrimination in both primary dataset (AUC=0.787, 95%CI:0.781 to 0.793;sensitivity=73.4%, specificity=72.2% ,positive predictive value=0.707,negative predictive value=0.868) and independent validation dataset (AUC=0.777, 95% CI:0.771 to 0.783,sensitivity=91.3% ,specificity=67.3% ,positive predictive value=0.607, negative predictive value=0.946). Conclusion The radiomics predictive model, which integrated with the radiomics signature and clinical characteristics can be used as a promising and applicable adjunct approach for preoperatively predicting the clinical stage (Ⅰ,Ⅱ stage vs. Ⅲ,Ⅳ stage) of patients with NSCLC.

Effect of image registration on free breathing MR diffusion kurtosis imaging in normal human kidney / 中华放射学杂志

Objective To investigate the effect of image registration on quantitative measurements of free breathing diffusion kurtosis imaging (DKI) in normal human kidney. Methods Twenty healthy volunteers were prospectively enrolled to undergo DKI imaging with a 3.0 T MR scanner. Three b values (0, 500, and 1 000 s/mm2) were adopted,with image registration performed after image acquisition. Acquired images were fitted using the DKI fitting model to generate the DKI metric maps,which were performed on both the pre-registration images and post-registration images. Image quality of the derived metric maps (before and after image registration,respectively) was assessed by two radiologists. Measurements of fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (D|), axial diffusivity (D⊥), mean kurtosis (MK), radial kurtosis (K|) and axial kurtosis (K⊥) were conducted. The inter-observer reproducibility of the image quality assessment was analyzed using intra-class correlation coefficient(ICC). Wilcoxon signed-rank test was used to evaluate the difference in the subjective scores of the metric maps between those obtained before registration and those after registration. While paired t test or Wilcoxon signed-rank test was performed to analyze the difference in the quantitative measurements of DKI metrics of the renal cortex and medulla between those obtained before registration and those after registration.Results For the inter-observer reproducibility, satisfactory ICCs were obtained for the quantitative metric measurements (pre-registration:0.784 to 0.821;post-registration:0.836 to 0.934). Significant difference was notice between subjective scores for the quality of metric maps (P<0.05 for each comparison). In both the renal cortex and medulla, significant difference was noticed between each metric value obtained with pre-registration images and that with post-registration images (P<0.05 for each comparison). Conclusion Image registration can not only offer higher quality DKI metric maps,but also has effect on the quantitative measurements of obtained metric maps.